Octopus Energy

Energy is a commodity market: customers cannot buy “better electrons,” so suppliers are judged on price, trust, billing accuracy and how they treat people when something goes wrong. Octopus had built its model around human service teams supported by the Kraken platform, which holds the entire customer lifecycle (account history, billing, meter data, communications) in one system (techUK, Aug 2024).

Then generative AI arrived in the middle of an energy crisis, when customer contact volumes were surging. Octopus began trialing AI on customer emails in February 2023 with a person supervising it; by the end of April it was answering 34% of all customer queries, the work of about 250 people in the UK alone, with an 80% customer satisfaction rating against 65% for humans, according to Greg Jackson (Sifted, 2024; Yorkshire Post, May 2023). That created the dilemma facing every service-heavy business: if AI answers routine customer emails faster, cheaper and sometimes better than people, do you remove the people, or change what human work is for?

Two paths were available. Path A: the AI-first cost-reduction route, would treat the “250 people’s work” as a savings opportunity: replace customer service staff with AI and book the headcount reduction. Path B: the route Octopus chose, would let AI absorb routine volume while humans kept ownership of review, judgment and the conversations where trust matters most, and the company kept growing.

The firm’s position was stated publicly at the moment the numbers came out:

A year later Jackson was framing the same strategy in explicitly human-amplifying terms: that generative AI has “turbocharged the power of the human” (Enlit World, June 2024).

The foundation is Kraken, the operating system Octopus originally built for itself. Kraken integrates the full customer lifecycle (account history, billing, meter readings, communications and product data) so service teams work from a single real-time view of the customer instead of fragmented tools (techUK, Aug 2024).

The generative AI layer is Magic Ink, Kraken’s customer service tool. Built on GPT-like models, it lets customer service staff summarize all interactions with a customer, generate draft responses, and suggest actions such as requesting a meter reading, drawing on continuous context from customer history and product specifications (techUK, Aug 2024).

The design is explicitly human-in-the-loop. To counter hallucinations, a verification system annotates verified facts with their sources and highlights text that could not be verified; team members are trained to review everything Magic Ink writes, and around a third of generated messages need zero to minimal changes before sending, meaning two thirds still get meaningful human editing. techUK states that “in all cases, the system keeps human operators in the loop” (techUK, Aug 2024). Octopus did not build an autonomous replacement for customer service; it built a drafting, summarizing and action-suggestion layer inside a human service model.

The same logic has since been extended well beyond the inbox. Kraken’s AI today coordinates smart tariffs and flexible assets: for example, over 160,000 customers charge their electric vehicles through Intelligent Octopus Go, the UK’s largest smart tariff, with Kraken coordinating the charging (Kraken case study). And in September 2025 Octopus announced the spin-off of Kraken as a standalone company: the platform was by then contracted to serve over 70 million household and business accounts worldwide, processing 15 billion new data points a day, with $500 million in committed annual revenue from licensing deals with energy majors including EDF, E.ON Next, National Grid US, Origin Energy, Plenitude and Tokyo Gas (Octopus Energy / BusinessWire, Sept 2025). In December 2025 the spin-off was completed with a $1 billion funding round led by D1 Capital Partners, valuing Kraken at $8.65 billion (Yahoo Finance / OilPrice, Dec 2025).

In each case the pattern is the same: the AI absorbs the repetitive, data-heavy coordination work, while humans keep ownership of judgment, service and the relationship.

Octopus Energy’s competitive moat is not only cheap energy, and not only software. It is the combination of software, AI and human service. Energy is a market where the core product is commoditized, so the difference lies in trust, customer experience, responsiveness and the ability to help people through complex or emotionally loaded situations.

The Hybrid Intelligence insight is precise: AI is structurally better at speed, scale, summarization, drafting, forecasting and pattern recognition across millions of accounts and data points. Humans remain structurally better at empathy, contextual judgment, accountability, vulnerability support and deciding what a customer actually needs when the situation is messy.

Octopus built AI to own the former so humans could focus on the latter. Magic Ink drafts and summarizes; Energy Specialists review, adapt and decide. Kraken automates much of the energy supply chain; human teams keep responsibility for service, trust and complex customer outcomes. AI forecasts energy flows and customer needs; humans design the products, relationships and interventions around that intelligence.

This is reskilling rather than simple automation. The routine email-answering workload was absorbed into an AI-assisted system, while human work moved toward higher-value service, complex customer care, technology, flexibility, field operations and engineering. Jackson’s framing that the “real focus is on upskilling and reskilling” is consistent with the company’s actual growth pattern: AI did not become a headcount-reduction lever; it became the operating layer that allowed the company to scale human service.

The contrast with a pure AI-first path is clear. Path A would have treated the 250-person equivalent as a cost-saving opportunity. Path B treated it as a chance to redesign the work: AI handles routine volume, humans handle trust and judgment, and the company scales both. That is why Octopus is a strong Hybrid Intelligence case: it shows that being highly AI-forward and highly human-centred are not opposites. In this case, they are the same strategy.